Activity War ROOM

Activity War ROOM is EXAWin's core operational screen. It records project-level meetings, calls, emails, and messages, then turns observed sales signals into P(Win), credible interval, impedance, momentum, and silence risk through the Bayesian engine.

The screen is not just a CRM activity log. It is designed to keep the operating loop connected:

Select project
→ Record meeting/activity
→ Enter attendees and observed signals
→ Update Bayesian posterior
→ Interpret the right-side analytics panel
→ Share the current view with the 1-Page Report when needed
→ Adopt system recommendations or link them to follow-up sales actions

Location: Sidebar → Sales Opportunity → Activity War Room

Screen Layout

Activity War ROOM uses three independent scroll panels. The project list, activity input/timeline, and project snapshot/analytics panel scroll independently.

When no project is selected, the center area shows an operating guide. Select a project from the left panel to activate the input form and analytics panel.

Left Panel: Project List

Filters and Search

Project Card Items

Closed projects are visually separated: Won uses emerald/green tones and Lost uses red tones.

Center Panel: Activity Input and Timeline

Center Header

For closed projects, a Project Closed notice appears at the top of the center panel. All activities remain viewable, but only the final closing activity can be edited.

Activity Input Form

The same form is used for new activity entry and activity editing.

Row 1: Basic Information

Past stages are disabled in gray and cannot be selected. Selecting a terminal stage such as Won or Lost closes the project; signal selection and Bayesian update are skipped for that terminal activity.

Row 2: Project Attendees / Stance

The current version uses project-stakeholder based attendance input, not a single main-contact selector.

This attendance and stance context becomes important meeting context for ontology interpretation and decision review.

Row 3: Observed Signals

Signals are the sales evidence observed in the activity. They are required for non-terminal stages.

Compound score is calculated by direction:

Compound = MAX(scores) + Σ(remaining scores) × 0.25

The strongest signal becomes the representative value, while the remaining signals contribute only 25% so a single activity does not create an unrealistic probability jump.

Row 4: Summary and Browser Local AI Assist

Summary is a required one-line title used on the activity card.

Below Summary, Browser local AI assist helps draft activity input. This is not server LLM or ontology reasoning. When supported by the browser, it summarizes user-entered content locally on the device.

Important: Local AI assist is an input helper. It never overwrites original text automatically. The user must explicitly apply a draft. Ontology-based interpretation and operational judgment are handled in the Ontology AI / Decision Console flow.

Row 5: Rich Text Three Tabs

Each tab uses a Quill rich text editor. The expand/collapse button adjusts editor height.

Row 6: Attachments

Attachments are collapsible. Click or drag files into the area. PDF, DOC, XLS, and image files are supported, with a 10 MB limit per file.

On Save

Save activity
├── Save as Draft
├── Save attendees and stance context
├── Save activity_signals relations
├── Preserve signal_data JSON snapshot
├── Create or recalculate BayesianUpdate for non-terminal stages
├── Save attachments
└── Record Audit Log

Activity Timeline

Activities are shown newest first. Each card includes activity number, date/time, type, duration, summary, stage, attendees, signals, meeting notes, action items, issues, attachments, comments, and likes.

Bayesian Badges

State Management

Draft activities can be confirmed with Confirm & Lock. Locked activities can be edited only after an Admin or project Owner unlocks them.

Closed Project Editing

Won/Lost projects cannot receive new activities. Only the final closing activity can be edited; changing its Stage back to the previous stage can reopen the project. Earlier activities are read-only.

Comments and Likes

Activity cards support comments, replies, and likes. Counts are aggregated back to the left project card. Comments can be deleted by the author or an administrator.

Right Panel: Project Snapshot and Analytics

The right panel appears on XL screens and above. Smaller screens prioritize the center timeline and input form.

Project Snapshot

Bayesian Analytics

Active Projects

Closed Projects

Closed projects show a terminal outcome hero first. The interpretive basis may be the final valid pre-terminal Bayesian snapshot rather than a new terminal calculation. The panel and report display a terminal note when this applies.

Recommendation Sales Action Status

History shows adoption time, linked activity, elapsed days, α/β impact, P(Win) impact, P-T impact, and decision impedance impact.

Confidence Band

Confidence Band shows the uncertainty around current P(Win): lower/upper 95% credible interval, P(Win) marker, standard deviation σ, evidence amount n, and margin. When evidence is weak, avoid making strong decisions from the headline probability alone.

Silence Risk

Silence Risk shows β risk accumulated from long gaps without contact.

Total silence risk = historical β penalty + current-gap expected β penalty

If there is no active penalty, the panel shows Stable. If a penalty exists, use the β increase and days since last activity to decide whether follow-up contact is needed.

Evidence & Trend

Evidence & Trend shows α/β accumulation, latest activity gains, and the P(Win) trend chart. When at least two updates exist, an SVG chart appears; hovering a point shows sequence, date, stage, P(Win), and CI range.

Quick Stats

Bayesian 1-Page Report

The current version connects Bayesian 1-Page Report directly inside Activity War ROOM. Open it from the center header or the right analytics panel.

The report opens as a modal, not a new page. Close it with the close button or ESC.

Report Structure

The report follows a posterior-first principle: posterior comes first; threshold, gap, and impedance are decision overlays above that posterior.

Operating Flow

General Activity Recording

1. Select a project from the left panel.
2. Click New or use the open input form.
3. Enter activity type, date/time, duration, stage, attendees, and stance.
4. Select observed signals for non-terminal stages.
5. Enter Summary manually or create a title draft with Browser local AI assist.
6. Record Meeting notes, Action items, and Issues.
7. Add attachments if needed.
8. Click Save activity.
9. Review the updated timeline and right analytics panel.
10. Open the 1-Page Report when the current interpretation needs to be shared.

Bayesian Processing After Save

Save activity
├── Calculate signal compound
├── Apply silence penalty to previous β when applicable
├── α_new = α_prev + SWV × Compound_positive
├── β_new = β_prev + SWV × Compound_negative
├── P(Win) = α_new / (α_new + β_new)
├── Calculate credible interval
├── Impedance = 1 / (1 + e^(-k × (P(Win) - T)))
└── Reflect in timeline / right panel / one-page report

Recommendation Follow-Up Linkage

1. Review the recommendation in the right analytics panel.
2. Adopt it as a sales action plan or mark it as not adopted.
3. If adopted, record the follow-up contact, preparation, or meeting result in Activity War ROOM.
4. When opened with a specific decision_id, the saved activity is linked as the omitted final meeting result for that recommendation.
5. Review linked activity, elapsed days, α/β impact, and P(Win) impact in the recommendation history.

Permissions

Data Relationships

Project
├── Customer / Owner / Main Contact
├── ProjectContact / Stakeholder stance
├── Activity
│   ├── ActivityAttendance
│   ├── ActivitySignal → SalesSignal → ImpactType
│   ├── BayesianUpdate
│   ├── ActivityAttachment
│   ├── Comment / Reply
│   └── Like
├── ProjectStage → SalesStage / SV / SWV / T / k
├── ProjectImpact → custom impact score
├── OntologyRecommendationLedger
│   ├── recommendation adoption state
│   ├── linked executed activity
│   └── baseline/current belief snapshot comparison
└── Bayesian 1-Page Report
    └── latest or pre-terminal Bayesian snapshot interpretation

Notes and Cautions

1. The core UX is the three independent panels. Use the left project list, center timeline/input, and right analytics panel independently.
2. Signal selection directly affects P(Win). Select only evidence actually observed in the activity.
3. Attendee and stance information matters. Record decision makers, influencers, supporters, and blockers as accurately as possible.
4. Browser local AI assist is only an input helper. Generated drafts are not saved automatically and must be reviewed before applying.
5. The 1-Page Report is a posterior-first interpretation tool. It summarizes the current snapshot but does not replace the original activity record.
6. Closed projects block new activities. Only the final closing activity can be edited, and changing its Stage back can reopen the project.
7. Confirm & Lock blocks edits and deletion. Ask an Admin or Owner to unlock if correction is needed.
8. Silence risk accumulates automatically. Review P(Win) and decision interpretation together when contact gaps become long.
9. Use recalculation carefully. Recalculate all Bayesian updates after Stage, Signal, Impact, or parameter changes.
10. Reports and analytics are evidence-based interpretation. Final sales judgment should also consider customer context, contract terms, and actual follow-up actions.